[Search for users] [Overall Top Noters] [List of all Conferences] [Download this site]

Conference orarep::nomahs::rdb_60

Title:Oracle Rdb - Still a strategic database for DEC on Alpha AXP!
Notice:RDB_60 is archived, please use RDB_70..
Moderator:NOVA::SMITHISON
Created:Fri Mar 18 1994
Last Modified:Fri May 30 1997
Last Successful Update:Fri Jun 06 1997
Number of topics:5118
Total number of notes:28246

5048.0. "RDMS$$EXE_CLOSE and COSI_MEM_GET_VM" by M5::DDAYBERR () Wed Feb 19 1997 16:10

      A customer is running RDB 6.0-15 and has a bugcheck
    with two exceptions.  The first exception- RDMS$$EXE_CLOSE
    + 5C and the second exception is  COSI_MEM_GET_VM +A7C.  
    My question is, which one ocurred first and which one is
    more indicitive of the actual problem?
    
         Also, the customer states that after moving her
    application off of a VAX system to an AXP system, that
    performance is worse.  Other than the logical area lock
    problem specific to axp's that was addressed in 6.0-16, is 
    there anything else we know of that might cause the problem.
    The customer claims the performance problem is definately 
    locking related.   Carry-over locks are disabled, as are 
    global buffers and snaps are non-deferred.  
    
                                  Debbie
T.RTitleUserPersonal
Name		DateLines
5048.1HOTRDB::LASTOVICAIs it possible to be totally partial?Wed Feb 19 1997 18:2215
    >    My question is, which one ocurred first and which one is
    >    more indicitive of the actual problem?
    
    	the exception that is 'furthest' from the top of the file is the
    first one that occured.  It is most likely the 'root' problem.  I'll
    have to assume (since you didn't mention) that the exception is an
    ACCVIO.  Exceptions in the memory management code are generally due to
    memory corruption of some sort; usually within Rdb.  We'd need to see
    the entire stack though.
    
    >performance is worse.
    
    	Is worse than what?  What sort of VAX and what sort of ALpha? 
    Disks, physical and logical database design?  Clustered?  What queries,
    more I/O, CPU?  What?
5048.2Thanks NormM5::DDAYBERRThu Feb 20 1997 16:1929
    
         The COSI_MEM_GET_VM was not followed by an accvio, but
    instead an internal consistency failure.  I'll see how big the
    bugcheck is and at least reply with the stack info - let me know
    what other specific information you want.
    
         Again, the customer states the specific performance problem
    is locking.  When they are experiencing the problem, all the users
    are hung, and we are having trouble tracking the real culprit.
    Looking at stall messages will show one waiter and 15-20 pages
    of blockers.   so they thenexecute rmu/show locks mode=blocking,
    but not all of the output is dumped - instead, then last message
    is "Additional blocking Processes not displayed".   
    
         So, we are having a difficult time determining who the 'top
    blocker' is.  Having the top blocker PID we could use SDA to see what
    the process has locked.
    
         We also know that most of the locking is on the root, I assume
    due to the fact that all the user apps are written to do EXCLUSIVE
    access.   So the amount of locks aren't in question, but there 
    seems to be a point where a top blocker hangs, and everyone else
    is hung behind him/her.    
    
        She will get the hardware information, but for the time being,
    I'm going to try to understand what causes the RDB processes to hang.
    
     
                                          Debbie